login
Search
Dave Korn announced GCC 3.4.6:
"This release is a minor release, containing fixes for regressions relative to earlier releases, but no new features. It is the final release from the 3.4.x series and the branch is now closed. It is thus also the final release from GCC series 3 overall."
GCC is the GNU Compiler Collection which includes C, C++, Objective-C, Fortran, Java, and Ada compilers. Download GCC 3.4.6 from a gcc.gnu.org mirror.
From: Dave Korn [email blocked] To: [email blocked] Subject: GCC 3.4.6 Released Date: Sun, 28 May 2006 07:02:25 +0100 GCC 3.4.6 was released on March 6, 2006. This release is a minor release, containing fixes for regressions relative to earlier releases, but no new features. It is the final release from the 3.4.x series and the branch is now closed. It is thus also the final release from GCC series 3 overall. This release is available from the FTP servers listed here: http://www.gnu.org/order/ftp.html The release is in the gcc/gcc-3.4.6 subdirectory. A list of known fixed bugs is available at http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 We would like to thank the vast number of contributors who have made this release possible, and who have made GCC series 3 the great success it has been. We look forward to working with you on GCC series 4. Thanks! -- Dave Korn, pp. Gabriel Dos Reis, GCC 3.4 Release Manager.
Wow!
End of an era, to be sure.
I remember when GCC was stuck in the 2.7.x blah-land. It inched along until EGCS lit a fire under it with what became 2.9x and later 3.0. That is what ultimately got GCC out of the mud and rolling with 3.x, and now w/ the Tree-SSA stuff in 4.x, GCC's got a full head of steam.
Here's a toast to just how far GCC's come along.
Now that we have final versio
Now that we have final versions for 2.x and 3.x, it would be nice to see a benchmark comparison between 2.95.6, 3.4.6, and 4.1.1. Maybe the SPECint suite w/ -O2.
People already did that. On a
People already did that. On average IIRC every next major version of GCC is 30-40% slower than previous.
Most performance is lost to code optimization, correctness and improved C/C++ support.
GCC guys were focused more in version 3.x on correct and full C/C++ implementation. GCC 4.x is focused on intergrating the optimization framework. I think the decision was right: (1) correctness is more important than compiler performance, (2) performance of the resulting code is more important than performance of the compiler itself.
Since the optmization work is finished, GCC devels promised that as of now no more features with negative compiler performance would be integrated and new focus would be performance of compiler itself.
@#$%^&*, I replied to wrong q
@#$%^&*, I replied to wrong question... I was talking about compiler itself...
Yeah. I don't really give a
Yeah. I don't really give a hoot how slow the compiler is; it's the produced code that matters.
Maybe too far, too fast
I think in some ways GCC is moving too fast. GCC 4.1 deprectes various silly compiler extensions from GCC 3.x (e.g. casted lvalues like 'void *a; (int *)a=b;'). That would be a welcome cleanup if it wasn't for the fact that the bootstrap compiler for GCC 3.3.x and earlier uses said extensions, meaning that the latest GCC can't self compile semi-recent versions of itself.
Hmm...
At least at one time, GCC had the goal of being bootstrappable by most vendor compilers as opposed to itself. Thus, GCC's source is supposed to use none of GCC's language extensions. (In fact, at one time, I believe they aimed at being compileable on K&R C compilers, but I think they gave up on that and just target ANSI C compilers. I could also be 100% mistaken. It's been awhile since I looked into it.)
I actually agree with deprecating certain language extensions. I've allowed some of those constructs to creep into my code over the years, and I get a rude shock when I try to compile that code on a stricter compiler. Cast-as-lvalue is one major example. In general, recasting pointers is just begging for non-portable code, so I can see this being on the list of things to chuck.
what about the qt-3.4.6 compilation bug?
As 3.x is closed now, is there any inofficial fix for that problem??
And I thought the open-source guys are better than MS. Just stopping support for the probably most heavily used compiler. :-(
http://bugs.gentoo.org/show_b
http://bugs.gentoo.org/show_bug.cgi?id=133301
Is it right that version
Is there something preventing you from accessing the front page of the website at gcc.gnu.org? Just wondering, since I can't possibly see any other legitimate reason for asking such a thing here.